Analytics
Building a Full-Funnel Attribution Model
No attribution model is true; each one answers a different question. How to choose models for each decision and combine them with tests you can trust.
BenchMarketing editorial team Updated October 2, 2026 3 min readShare
Attribution tries to answer a simple question: which marketing caused this sale? The honest answer is that no model knows. Each model is a rule for sharing credit, and different rules suit different decisions.
The common models
- Last click: all credit to the final touch before conversion. Simple, and biased toward search and retargeting, which sit closest to the purchase.
- First click: all credit to the first touch. Useful for asking which channels introduce new customers; ignores everything after.
- Linear and position-based: credit spread across touches by fixed rules. More balanced, but the weights are arbitrary.
- Data-driven: credit assigned by a model trained on your conversion paths. Google Ads and GA4 both offer one. Better than fixed rules, but it only sees the touches its platform can track.
Why platform attribution disagrees
Each ad platform sees its own ads and little else. Meta does not know a customer also clicked a Google ad; Google does not know they watched a TikTok video. Each platform credits itself, and the totals overlap. That is not a bug to fix but a limit to work around.
A full-funnel approach that holds up
- Agree on one source of truth for revenue: your store, CRM or finance system, not an ad platform.
- Track blended efficiency: total revenue against total spend (MER). It cannot double count.
- Use platform attribution for decisions inside a platform: which campaign, ad set or creative to scale.
- Use a neutral analytics tool for cross-channel paths: GA4 or a similar tool with consistent UTM tagging, to see how channels combine.
- Test incrementality for big budget questions: pause or hold out a channel in some regions or for some audiences and measure the change in total sales.
- Add marketing mix modelling when spend is large enough: it estimates each channel's contribution from spend and sales over time. See MMM for beginners.
Make the inputs clean
Attribution is only as good as the tracking under it. Tag every link with consistent UTM parameters (the UTM builder helps), deduplicate conversions sent from both browser and server, and keep conversion windows consistent when comparing channels.
Compare models side by side with the attribution model comparator.
About the figures
Benchmark figures in this article come from the BenchMarketing dataset and update when the benchmark pages do. Each benchmark page lists its sources and period; see our methodology.
Sources
See where you stand
Compare your own numbers with these benchmarks for your industry.
Create a free account